Analysis of N-Glycoproteins Using Genomic N-Glycosite Prediction

Oct 29, 2013 - Analysis of N-Glycoproteins Using Genomic N-Glycosite Prediction ... Memorial Sloan-Kettering Cancer Center, New York City, New York ...
0 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/jpr

Analysis of N-Glycoproteins Using Genomic N-Glycosite Prediction Shisheng Sun,†,# Bai Zhang,†,# Paul Aiyetan,† Jian-Ying Zhou,† Punit Shah,† Weiming Yang,† Douglas A. Levine,‡ Zhen Zhang,† Daniel W. Chan,† and Hui Zhang*,† †

Department of Pathology, Johns Hopkins University, Baltimore, Maryland 21287, United States Memorial Sloan-Kettering Cancer Center, New York City, New York 10065, United States

‡

S Supporting Information *

ABSTRACT: Protein glycosylation has long been recognized as one of the most common post-translational modifications. Most membrane proteins and extracellular proteins are N-linked glycosylated, and they account for the majority of current clinical diagnostic markers or therapeutic targets. Quantitative proteomic analysis of detectable N-linked glycoproteins from cells or tissues using mass spectrometry has the potential to provide biological basis for disease development and identify disease associated glycoproteins. However, the information of low abundance but important peptides is lost due to the lack of MS/MS fragmentation or low quality of MS/MS spectra for low abundance peptides. Here, we show the feasibility of formerly N-glycopeptide identification and quantification at MS1 level using genomic N-glycosite prediction (GenoGlyco) coupled with stable isotopic labeling and accurate mass matching. The GenoGlyco Analyzer software uses accurate precursor masses of detected N-deglycopeptide peaks to match them to N-linked deglycopeptides that are predicted from genes expressed in the cells. This method results in more robust glycopeptide identification compared to MS/MS-based identification. Our results showed that over three times the quantity of N-deglycopeptide assignments from the same mass spectrometry data could be produced in ovarian cancer cell lines compared to a MS/MS fragmentation method. Furthermore, the method was also applied to N-deglycopeptide analysis of ovarian tumors using the identified deglycopeptides from the two ovarian cell lines as heavy standards. We show that the described method has a great potential in the analysis of detectable N-glycoproteins from cells and tissues. KEYWORDS: glycosylation, glycosite prediction, genome-wide, SILAC, accurate mass matching, ovarian cancer, mass spectrometry

■

INTRODUCTION N-Linked glycoproteins play important roles in biological processes, including cell-to-cell recognition, growth, differentiation and programmed cell death, viral evolution. and immune escape.1−5 Specific N-linked glycoprotein changes are associated with disease progression, and identification of these N-linked glycoproteins has a potential application in disease diagnosis, prognosis, and prediction of treatments.6,7 Tandem mass spectrometry (MS/MS)-based shotgun proteomics technology combined with stable isotope dilution has become an effective approach for large-scale protein identification and quantification in complex biological or clinical samples.8−11 A typical shotgun proteomic analysis consists of digestion of proteins to peptides and analyzing the peptides by liquid chromatography−tandem mass spectrometry (LC−MS/MS). Prior to LC−MS/MS analysis, digested peptides are optionally labeled with isotopic or isobaric tags for peptide and protein quantitation.10,11 Alternatively, proteins are metabolically labeled by stable-isotope labeling by amino acids in cell culture (SILAC) (typically lysine and/or arginine) before they are digested to peptides for mass spectromety analysis.12 Each tandem spectrum is searched through the database against all possible peptides and a peptide sequence based on the highest correlation of theoretical MS/MS spectrum to the acquired MS/MS spectrum to assigned to © 2013 American Chemical Society

the spectrum. The assigned peptide is then quantified by the number of spectra assigned to each peptide or by using the isotope or isobaric tags for accurate quantification. The process is very convenient since many database search engines and software have been developed to allow the automated assignment of MS/MS spectra to the peptide sequences.13−15 This workflow for the identification of specific peptide relies on two factors: (1) MS/MS spectra are generated from all peptides by mass spectrometer, and (2) MS/MS spectra are in high quality to generate enough information for correlation to the theoretical spectrum. However, some peptides have inherent sequences that do not generate high quality MS/MS, and precursor ions of low abundance peptides are either not selected for MS/MS acquisition or produce poor quality MS/ MS spectrum. These factors have greater effect on the identification of protein modifications than the identification of the protein, as a protein can be identified by multiple peptides, while the modifications can be identified only when the modified peptide is identified. Meanwhile, LC−MS data contain a lot more information on the parent ions of peptides (accurate mass, retention time, abundance, etc.), and these have been employed for peptide identification.16,17 However, in this Received: June 17, 2013 Published: October 29, 2013 5609

dx.doi.org/10.1021/pr400575f | J. Proteome Res. 2013, 12, 5609−5615

Journal of Proteome Research

Article

0.5% SDS/1 M NH4HCO3 buffer directly. Lysates were briefly sonicated till the solutions were clear. Protein concentrations were determined by BCA protein assay reagent (Pierce, Rockford, IL).

approach, a database with a list of identified peptides by MS/ MS is established prior to the peptide identification using LC− MS data (e.g., AMT tag database based on MS/MS identification) due to the high complexity of proteomes in biological and clinical samples. Unlike global proteomics, N-linked glycoproteomics focuses on the known, well-defined N-linked glycopeptides.18,19 The Nglycosylation of proteins occurs at N-X-S/T motif (where X is any amino acids except proline), and thus the potential Nglycopeptides from a genome or proteome can be predicted from genomic data by containing the N-linked motifs.20,21 The number of N-glycopeptides in this case is significantly reduced compared to the tryptic peptides increasing the potential for identification based on their distinct masses. The high throughput and accuracy of SILAC labeling has made it a widely used method in quantifying proteomes of cells, tissues, and even blood.12,22−25 The isotopic pairs and lysine and arginine count (K/R count) information containing in the mass difference between light and heavy SILAC peptides have the potential to enhance the identification accuracy of Ndeglycopeptide based on accurate mass matching to predictive expressed peptides. In this study, we describe a new method, named GenoGlyco, to show the feasibility of quantitative N-glycoproteome analysis based on genomic N-glycosite prediction. The peptides that are N-linked glycosylated in the native proteins were selectively isolated using solid-phase extraction of N-linked glycopeptides (SPEG), which is based on hydrazide chemistry,18,26 and the deglycosylated forms of these peptides from metabolic labeled cells were analyzed by liquid chromatography−mass spectrometry (LC−MS) and tandem mass spectrometry (MS/MS). The N-deglycopeptide pairs were identified by accurate mass matching of isolated peptides to the N-glycosites from the genome-wide prediction of the expressed glycoproteins in the cells (GenoGlyco). The SILAC peptide peak pairs containing K/R count information allow the accurate measurement of the deglycopeptide from potential deglycopeptides with the same mass. We showed that this method allowed the assignments of 1772 deglycopeptides detectable by LC−MS using accurate mass mapping of detected peptide peaks in the SKOV-3 cell line. Among these, 501 deglycopeptides were verified by the traditional shotgun approach using MS/MS spectra and database search, and an additional 184 deglycopeptides were verified using at least 3 fragment ions from MS/MS spectra. The SILAC-labeled and assigned deglycopeptides were used as standards for the N-glycoprotemic analysis of ovarian cancer tissue.

■

Ovarian Carcinoma Tissue

An optimal cutting temperature-embedded (OCT) frozen highgrade serous ovarian cancer tissue was obtained from Memorial Sloan-Kettering Cancer Center, New York. The tissue was cut into small pieces and lysed by 8 M urea/0.5% SDS/1 M NH4HCO3 buffer as described for cell lines. Formerly N-Glycopeptides Isolation

For each cell line, the N-deglycopeptides were isolated from both light cell proteins and protein mixtures of the same amount of light and heavy cell proteins. For tissue samples, the SILAC-labeled cell proteins from SKOV3 and OVCAR-3 were first mixed together with same amount to prepare SILAC mix. The same amount of SILAC cell protein mix was added into each tissue samples as internal standards. Then N-deglycopeptides were isolated from whole cell extracts of both cells and tumor tissue sample directly by employing solid-phase extraction of N-glycopeptides method as described previously with minor modifications.26 Briefly, 1 mg of proteins was reduced by 5 mM DTT at 37 °C for 1 h and alkylated by 20 mM iodoacetamide at room temperature in the dark for 30 min. After the solutions were diluted 8-fold with 0.1 M NH4HCO3 buffer, and 20 μg of trypsin (Promega, Madison, WI) was added and incubated at 37 °C overnight with shaking. Samples were centrifuged at 13,000g for 10 min to remove any particulate matter and cleaned by C18 column (Milford, MA, Waters). Peptides were eluted by 400 μL of 60% ACN/0.1% TFA. Glycopeptides were oxidized by 10 mM NaIO4 solution at room temperature for 1 h in the dark; then the samples were cleaned by C18 column again. The eluted solution was collected into 25 μL of equilibrated hydrazide beads (BioRad, Richmond, CA) directly and incubated with 100 mM aniline at room temperature for 3 h.27 The beads were washed 3 times each with 50% ACN, 1.5 M NaCl, water, and PBS buffer. N-Deglycopeptides were released via 3 μL of PNGase F (New England Biolab, Beverly, MA) in PBS buffer at 37 °C overnight with shaking. N-Deglycopeptides were collected in supernatants and wash solutions and cleaned by C18 column. The N-deglycopeptides were dried via SpeedVac and resuspended in 20 μL 0.1% TFA solution for mass spectrometry analysis. Mass Spectrometry Analysis

All deglycopeptide samples from light cells, mixed cells (H:L = 1:1) and tissues (H:L = 1:1) underwent 3 replications of LC− MS/MS analysis for one biological and technical replicate of each sample with ∼1 μg (2 μL) per run. Peptides were separated on a Dionex Ultimate 3000 RSLC nano system (Thermo Scientific, Bremen, Germany) with a 75 μm × 15 cm Acclaim PepMap100 separating column (Thermo Scientific) protected by a 2 cm guarding column (Thermo Scientific). Mobile phase flow rate was 300 nL/min and consisted of 0.1% formic acid in water (A) and 0.1% formic acid 95% acetonitrile (B). The gradient profile was set as follows: 4−35% B for 70 min, 35−95% B for 5 min, 95% B for 10 min, and equilibration in 4% B for 15 min. MS analysis was performed using an Orbitrap Velos Pro mass spectrometer (Thermo Scientific). The spray voltage was set at 2.2 kV. Orbitrap spectra (AGC 1x106) were collected from 400 to 1800 m/z at a resolution of

MATERIALS AND METHODS

SILAC Labeling of SKOV-3 and OVCAR-3 Cells

Human ovarian carcinoma cell lines SKOV-3 (ATCC, HTB77) and OVCAR-3 (ATCC, HTB161) were obtained from ATCC (Rockville, MD). Both cell lines were cultured in SILAC RPMI 1640 medium containing heavy isotope-labeled 13C6-L-lysine and 13C615N4-L-arginine (Cambridge Isotope Laboratories, Andover, MA) supplemented with 10% dialyzed fetal bovine serum (diFBS) (Invitrogen, Carlsbad, CA) or in normal RPMI 1640 medium with 10% FBS to generate the SILAC-labeled and normal cell proteins, respectively. Cells were cultured for approximately 10 doublings in the SILAC medium to ensure complete labeling. After removal of the medium, the cells were washed 3 times with PBS buffer and then lysed by 8 M urea/ 5610

dx.doi.org/10.1021/pr400575f | J. Proteome Res. 2013, 12, 5609−5615

Journal of Proteome Research

Article

Figure 1. Workflow of the genomic N-glycosite prediction (GenoGlyco) method for glycoproteomic analysis.

specific N-deglycopeptide based on the monoisotopic mass of the light peptides with 10 ppm mass error tolerance, and then the K/R count of the peak pairs was employed to filter the results. The matched unique or multiple peptides to each SILAC peak pair with the same mass and K/R count were considered as the N-deglycopeptides or N-deglycopeptide candidates of the peak pair. The replicates of deglycopeptides resulting from different runs or different charge states of the same peptide within 2 min retention time range were then removed. The script for this analysis was written and run in the R programming environment.

60K followed by data-dependent HCD MS/MS (at a resolution of 7500, collision energy 45%, activation time 0.1 ms) of the 10 most abundant ions using an isolation width of 2.0 Da. Charge state screening was enabled to reject unassigned and singly charged ions. A dynamic exclusion time of 35 s was used to discriminate against previously selected ions. Database Search

All LC−MS/MS data were searched against corresponding cell expression databases by both Sequest (proteome discoverer) and MaxQuant (v1.3.0.5). The search parameters for Sequest were set as follows: up to two missed cleavage were allowed for trypsin digestion, 10 ppm precursor mass tolerance and 0.06 Da fragment mass tolerance; carbamidomethylation (C) was set as a static modification, while oxidation (M), deamination (N), 13 C6 label of lysine (K), and 13C615N4 arginine (R) were set as dynamic modifications; SILAC 2plex (Arg10, Lys6) was selected as the quantification method, and the results were filtered with 1% FDR. For MaxQuant, the search parameters were as follows: two missed cleavages were allowed for trypsin digestion. Carbamidomethylation (C) was set as static modification, and oxidation (M) and deamination (N) were set as dynamic modifications. Two multiplicities with “Arg10” and “Lys6” were selected as heavy labels. 1% FDR was employed for peptide identification, and all peptides were used for protein quantification. All other settings were set as default values.

■

RESULTS AND DISCUSSION

GenoGlyco Method for N-Linked Glycoproteomics

Here we describe the GenoGlyco method for quantitative Nglycoproteome analysis based on genome-wide glycosite prediction coupled with accurate mass measurement of SILAC-labeled deglycopeptide pairs (Figure 1). The GenoGlyco method in this paper includes the following steps: (1) NDeglycopeptides were specifically isolated from light and heavy SILAC-labeled cell proteins by the solid-phase extraction of Nlinked glycopeptides (SPEG) based on hydrazide chemistry. (2) The accurate mass of the mixture of light and heavy labeled deglycopeptides was measured by LC−MS. (3) Potential Ndeglycopeptides from the SILAC-labeled cell lines were predicted using the N-glycosylation motif and the gene expression data of the cell. (4) The deglycopeptide peak pairs were matched to predicted deglycopeptides of the cell line using accurate mass matching, and the K/R count in the peptides was derived from the mass difference of SILAC pairs. We verified the assigned N-deglycopeptides by MS/MS. We showed that the heavy SILAC-labeled N-deglycopeptides identified using this method could be used for the analysis of ovarian cancer tissues.

Identifying Formerly N-Glycopeptides by GenoGlyco Method

GenoGlyco Analyzer is an algorithm that matches mass spectrometric peak masses to those predicted deglycopeptide masses from a sample. GenoGlyco Analyzer predicts potential N-deglycopeptides with consensus N-linked glycosylation sequons (N-X-S/T, where X is not P) in a sample and matches the experimental detected peaks (or peak pairs) by mass spectrometer to the predicted deglycopeptides using accurate mass matching and the K/R count information based on the mass difference if the peptides are derived from SILAC-labeled samples. Potential N-deglycopeptides of cells were predicted from the cell expression database28 by searching N-X-S/T (X cannot be proline) motifs (Supporting Information). The SILAC peak pair information of N-deglycopeptides of both cell lines was extracted from raw MS data into an ‘allpeptides.txt’ file when searching the database using MaxQuant as described above. The SILAC peak pairs were matched to the cell line

Prediction of Potential N-Glycopeptides from a Cell Line Using N-Glycosylation Motif and the Gene Expression Data

First, we determined the theoretical feasibility of the GenoGlyco method by employing the SKOV3 ovarian cancer cell line as a model. We created a SKOV3 cell expressed protein database based on mRNA microarray data from the cell-miner repository developed by the NCI genomics and bioinformatics group28 and human IPI 3.87 database29 and then predicted all potential N-glycopeptides using the gene expression data from the cells. Compared to the tryptic peptides in the entire IPI 5611

dx.doi.org/10.1021/pr400575f | J. Proteome Res. 2013, 12, 5609−5615

Journal of Proteome Research

Article

database (3,565,543 unique tryptic peptides with up to 2 missed cleavage sites), only ∼4.5% of the tryptic peptides (159,028 peptides) were potential N-glycopeptides predicted by consensus N-X-T/S motif (X denotes any amino acids except proline)20 and gene expression data in SKOV3 cells (14,652 genes).28 Among 159,028 potential N-linked glycopeptides containing the consensus N-X-T/S motif from peptides from SKOV3 cells, about one-third (∼50,000 with up to 2 missed cleavage sites) are from transmembrane proteins, cell surface proteins, or secreted proteins that were potentially N-glycosylated.21 These results showed that the targeted analysis of N-deglycopeptides dramatically reduced the complexity of the peptides over the total tryptic peptides in the protein database (Table 1).

addition, due to the different mass shifts introduced by SILAC labeling for the K and R containing peptides, the mass difference of the paired light and heavy labeled peptides could be used to determine the number of K or R in the peptides. The K/R count information contained in SILAC peak pairs makes 91.4% of N-deglycopeptides (145,384) unique within 1 ppm mass error and 65% of N-deglycopeptides (103,357) unique within 10 ppm mass error (Figure 2A). The reduced number of potential N-deglycopeptides and the K/R count information in SILAC peak pairs dramatically increased the unique N-deglycopeptides for each distinct mass detected by mass matching using LC−MS spectra.

Table 1. Number of Proteins and Potential NDeglycopeptides Expressed in SKOV3 and OVCAR-3 Cell Lines

Next, we assessed the feasibility of the GenoGlyco method by applying it toward the identification of N-deglycopeptides from SKOV3 cells. SKOV3 cells were labeled by light and heavy SILAC. Equal amounts of proteins from labeled cells were mixed, and the N-deglycopeptides were isolated by solid-phase extraction of N-glycopeptides (SPEG) and analyzed by triplicate LC−MS using orbitrap velos mass spectrometer. A total of 8621 peak pairs (average 2874 per run) were detected from SKOV3 cells with a mass accuracy of 10 ppm. Out of these peak pairs, 3218 pairs contained only K (mass difference of 6, 12, or 18 Da), 3439 pairs contained only R (mass difference of 10, 20, or 30 Da), and 1964 pairs contained both K and R (mass difference of 16, 22, or 26 Da). After matching the light mass of all peak pairs with potential N-deglycopeptides of the proteins expressed in SKOV3 cells using K/R count information with10 ppm mass error, we were able to assign 1,772 unique peak pairs with a specific mass and retention time to N-deglycopeptides, 64.3% (1,139) peak pairs matched to unique N-deglycopeptide mass, and 97.3% peak pairs (1,725) matched to no more than three deglycopeptides (Figure 2B and Supplementary Table S1). The results showed that most of the N-deglycopeptides (∼2/3 of identified deglycopeptides) from a cell line could be distinguished and assigned by their mass alone within 10 ppm mass error. These are consistent with results from the predicted N-deglycopeptides. The separation will be increased when mass accuracy reaches 1 ppm (Figure 2A).

database

proteins

peptidesb

N-deglycopeptidesc

HUMAN IPI3.87 Human RefSeq SKOV3 cell proteins OVCAR3 cell proteins

91,491 36,430 14,652 14,888

3,565,543 2,884,481 1,446,496 1,471,285

360,665 298,511 159,028 162,177

Identification of the SILAC-Labeled N-Deglycopeptide Pairs

a

Human IPI3.87 protein database were downloaded from the IPI database29 (total 91,491 protein entries). Human Refseq protein database was downloaded from NCBI Web site31 on July 29th, 2013. The expressed protein database for SKOV-3 and OVCAR-3 cell lines were derived from RNA expression data.28 bNumber of tryptic peptides with up to 2 missed cleavage sites. cNumber of tryptic peptide containing N-X-S/T motif (X is any amino acids except Pro) with up to 2 missed cleavage sites.

The reduced complexity of potential N-deglycopeptides reduced the mass overlap of peptides and allowed the separation of these deglycopeptides by their mass. We calculated the number of unique deglycopeptides to each mass. With a high mass accuracy of 1 ppm, which is achievable using the current mass spectrometer, over 65% of potential Ndeglycopeptides (104,748) contained distinct masses. With a decrease of instrument mass accuracy (10 ppm), the percentage of unique N-deglycopeptides with distinct masses was reduced to 13.6% (21,575) of all potential N-deglycopeptides. In

Figure 2. Feasibility of the glycoprotein identification in the SKOV-3 cell line using the GenoGlyco method. (A) Theoretical analysis of potential Ndeglycopeptides with distinct mass with or without K/R count information in the SKOV-3 cell line. (B) Analysis of N-deglycopeptides isolated from SKOV-3 cells using the GenoGlyco method. 5612

dx.doi.org/10.1021/pr400575f | J. Proteome Res. 2013, 12, 5609−5615

Journal of Proteome Research

Article

Figure 3. Analysis of N-deglycopeptides identified in the SKOV-3 cell line by GenoGlyco and MS/MS-based identification. (A) Specificity of the isolated N-deglycopeptides identified by LC−MS/MS and database search. (B) Examination of identified 1,772 N-glycosites by GenoGlyco. A total of 503 N-glycosites were verified by MS/MS identification. Of the remaining 1,269 N-glycosites, 781 were not subjected to data-dependent MS/MS analysis during the same LC−MS-MS/MS analysis, but an additional 3 LC−MS/MS runs of the light N-deglycopeptide samples identified and verified 103 of these. MS/MS spectra were generated for the remaining 488 SILAC labeled deglycopeptide pairs. Of these, 81 MS/MS spectra contained at least 3 fragment ions matched to theoretical fragment ions using targeted spectrum investigation, and the remaining 407 deglycopeptides were of low quality MS/MS spectra.

Verification of Identified N-Deglycopeptides by Tandem Spectra

assignment of MS/MS spectra to peptide sequence by database search algorithm still failed. Third, low quality of MS/MS spectrum caused by high background f ragment ions. These spectra (407 peak pairs) were mostly generated from low abundant MS peaks with high background MS/MS ions. The quality of MS/ MS spectra are known to be related to the abundance of peptides, and therefore, low abundant deglycopeptides can be only detected by MS but are not able to generate quality MS/ MS spectra for sequence assignment. Our results showed that assigning N-deglycopeptide sequences by GenoGlyco and mass spectrometric analysis of N-deglycopeptides reduces the obstacles of traditional shotgun proteomics and allows the assignment of detectable N-deglycopeptides from samples. Examination of precursor ions of 58 MS/MS only deglycopeptides showed that 19 identified N-deglycopeptides contained other modifications (deamidation), two contained more than three K/R, and one contained both light lysine and heavy arginine (false identification). Sixteen N-deglycopeptides peak pairs did not contain K/R count information (isotopic type) in MaxQuant results, and 13 N-deglycopeptides were not detected as pairs, while another 6 peak pairs contained K/R counts different from the identified ones (Supplementary Table S3). The undetected peak pairs and the deviation of H/L ratios of many deglycopeptides (Supplementary Tables S1, S2, and S3) in the study might be caused by the different components in normal and SILAC medium as well as normal and dialyzed FBS used for cell culture in this study. This situation should be improved by using same medium and dialyzed FBS with normal and heavy isotope labeled amino acids for culturing light and heavy SILAC cells, respectively. Reducing the complexity by capturing N-deglycopeptides is a prerequisite for the identification accuracy of the method. The database of potential deglycopeptides expressed from cell lines contains only 4.5% of unique peptides in the entire human database. SILAC labeling and detecting peptides by peak pairs in LC−MS help the detection of these peptides. Additional peaks present in MS spectra are due to chemical noise and other contamination, but metabolic labeling of cell glycopeptides with SILAC peak pairs can exclude these contaminants. Moreover, the MS/MS-based identification indicated that all methionine residues in identified N-deglycopeptides isolated by the SPEG method were in the oxidized state, which might be due to the oxidation of sodium periodate.30 All of these reduced the complexity of deglycopeptide samples and thus enhanced the identification accuracy of deglycopeptides at the MS1 level. The specificity of the method can be further enhanced by an

To determine the specificity of the identified N-deglycopeptides using the GenoGlyco method, we evaluated the assigned deglycopeptides using the MS/MS spectra collected from the same raw files. The MS/MS spectra were searched against the SKOV3 cell expressed protein database using Sequest13 with up to 2 missed tryptic cleavage sites and 10 ppm mass error. The assigned peptide sequences using MS/MS spectra (5565 spectra with 1% FDR rate) showed that 93.2% (561 unique peptides) of assigned unique sequences contained consensus N-linked glycosylation motif, showing high specificity of the isolated deglycopeptides from ovarian cancer cells (Figure 3A and Supplementary Table S2). Among the deglycopeptides identified by MS/MS spectra, 503 of them (89.7%) were included in the 1,772 matched deglycopeptides and could be used to verify the results of deglycopeptide assignment. (Figure 3B and Supplementary Table S1). We further investigated the 1269 peak pairs that were assigned by GenoGlyco but not identified by MS/MS spectra and the database search (Figure 3B). Examination of the MS and MS/MS spectra of these peak pairs showed three reasons for the failure of identification of observed MS peaks by MS/ MS and database search using MS/MS spectra. First, a lack of MS/MS spectra due to the complexity of the samples. The majority of peak pairs (781) were not identified due to the lack of MS/ MS acquisition for the peak pairs. Analysis of these peptides by three additional LC−MS/MS spectra of N-deglycopeptides allowed the identification of an additional 103 peaks (Figure 3B and Supplementary Table S1). Second, low number of f ragment ions f rom the selected peptides. The remaining 488 peak pairs were selected by data-dependent acquisition for tandem mass spectrometry. However, the quality of the MS/MS spectra was not informative enough to allow the assignments of the peptide sequences from these MS/MS spectra using the database search algorithm. Among the 488 peak pairs with MS/MS spectra, we identified 81 peak pairs containing at least 3 fragment ions correctly assigned to the fragment ions from the predicted Ndeglycopeptide sequences identified by GenoGlyco (Figure 3B). These peptide sequence−dependent MS/MS spectra may cause the lack of identification of these peptides. Additional evidence for the failure of peptide identification from these peptides caused by their sequence-dependent fragmentation pattern came from the analysis of these peptides by 2D-LC− MS/MS. The data showed that similar spectra were generated from these peptides by increased level of these peptides, but the 5613

dx.doi.org/10.1021/pr400575f | J. Proteome Res. 2013, 12, 5609−5615

Journal of Proteome Research

Article

additional 2 mass units by incorporating 18O mass tags on Nglycosites of formerly heavy N-glycopeptides during the removal of N-glycans by PNGase F. The 18O mass tag can provide the glycosite count information in SILAC peak pairs and distinguish the deglycopeptide from non-glycopeptide peak pairs. In addition, our theoretical data showed that the GenoGlyco method also works when the potential deglycopeptide database was generated from the human entire protein database (RefSeq as of July 29, 201331) instead of the cell expressed database and deamidation was set as variable modification from additional potential N-glycosites if a peptide contains more than one N-XS/T motifs. 87.7% (298806) and 49.5% (168,630) of potential N-deglycopeptides contained distinct masses within 1 ppm and 10 ppm, respectively (Supplementary Figure S1).

We then pooled the heavy labeled N-deglycopeptides extracted from both SILAC-labeled ovarian cancer cell lines and made SILAC N-deglycopeptide mix as heavy isotopelabeled standards with 4075 nonredundant assigned Ndeglycopeptides from the two cell lines. The mix of SILAClabeled N-deglycopeptides from both cells increases the number of N-deglycopeptide standards for analysis of ovarian tumors (Figure 4A). The SILAC-labeled N-deglycopeptide standards were used to analyze ovarian tumor. Tumor tissues are often stored frozen embedded in the optimal cutting temperature (OCT) medium. We applied the SILAC-labeled N-deglycopeptide standards to analyze the OCT embedded ovarian cancer tissues. The SILAC deglycopeptide standards were added to N-deglycopeptides isolated from ovarian tumor stored frozen with OCT embedding. All detected peak pairs were matched with the assigned SILAC N-deglycopeptides based on mass and the K/R count of the deglycopeptides. We were able to identify 835 N-deglycopeptides in OCT-embedded tumor tissues (Figure 4B, Supplementary Table S6); 326 assigned peak pairs were validated via MS/MS-based identification (Figure 4B and Supplementary Tables S6 and S7). Additionally 509 N-glycosites were identified by SILAC peak pair mapping (Figure 4B). Furthermore, MS/MS-based method identified an additional 396 N-glycosites: 319 of them were identified only in tissues and most of them did not originate from ovarian cancer cells but rather from plasma glycoproteins (Figure 4B and Supplementary Tables S6 and S7). These results demonstrated that hundreds of N-deglycopeptides from tumor tissues could also be analyzed by LC−MS-MS/MS using the GenoGlyco method.

Analysis of Ovarian Cancer Tissue Using SILAC-Labeled N-Deglycopeptides from Multiple Ovarian Cancer Cell Lines

The identified N-deglycopeptides from SILAC labeled cells could be used for the analysis of N-deglycopeptides from a variety of biological or clinical samples such as cells, tissues, and body fluids (Figure 1). In this study, we used the SILAClabeled N-deglycopeptides from ovarian cancer cells as universal standards for the analysis of ovarian tumor. To increase the coverage of SILAC-labeled N-deglycopeptides as standards for ovarian tissue analysis, we applied the same procedure to the analysis of another ovarian cancer cell line, the OVCAR-3 cell line. Theoretically, the potential Ndeglycopeptides from OVCAR-3 cells (162,177 potential Ndeglycopeptides with 2 missed cleavage sites) also contain 4.5% of tryptic peptides from human IPI database (Table 1). The unique N-deglycopeptides with distinct peptide mass from potential N-deglycopeptides increased from 13.4% by mass matching only to 64.6% by both mass and K/R count information with mass tolerance of 10 ppm (Supplementary Figure S2). From the LC−MS analysis of SILAC-labeled OVCAR-3 cells, we found 60.4% (1055) peak pairs were assigned to unique N-deglycopeptides when coupled with K/R information (Supplementary Figures S3, S4-A). Among 1748 peak pairs assigned to N-deglycopeptides using SILAC peak pairs, 436 of them were also identified by Sequest at 1% FDR (Supplementary Tables S4 and S5 and Figure S4-B). Another 133 peak pairs were verified by MS/MS spectra with at least 3 fragment ions, and 168 peak pairs were verified by three additional LC−MS/MS spectra of light N-deglycopeptides (Figure 4B and Supplementary Table S4). These results indicated that assigning SILAC peak pairs of N-deglycopeptides to unique N-deglycopeptide sequences should be applicable to most cell lines.

■

CONCLUSIONS We described the theoretical feasibility of N-glycoprotein analysis using genome-wide mapping of N-glycosites and tested the feasibility by mapping the deglycopeptide peak pairs from SILAC-labeled ovarian cancer cell lines to identify N-glycosites using both accurate mass and K/R count information. By employing the GenoGlyco Analyzer software, which identifies the N-deglycopeptides by accurate mass matching of precursor masses to that of predicted N-deglycopeptides, we identified more than three times the quantity of N-deglycopeptides in two ovarian cancer cell lines compared to a MS/MS fragmentation based identification method. The identification specificity and accuracy of this method can theoretically be further increased by incorporating 18O on N-glycosites of N-deglycopeptide and increasing mass accuracy of SILAC-labeled lysine/arginine peptides. This method has a potential to identify all detectable N-deglycopeptides and N-glycoproteins from a cell line or even tissue and promises to be helpful in biomarker discovery. By applying super SILAC-labeled N-deglycopeptides for each clinical specimen, this method can be used to identify and quantify thousands of N-deglycopeptides from clinical specimens of a small size (1 mg). This is especially useful for the proteomic analysis of some important specimens with limited available materials. The candidate glycoproteins identified from clinical samples using this method can be further validated by independent methods, such as applying a MudPIT for peptide separation and MS/MS identification, targeted MS/MS analysis, MRM experiment, and Western blot.

Figure 4. Identification of N-deglycopeptides in ovarian cancer tissue by GenoGlyco using SILAC-labeled deglycopeptides from multiple ovarian cancer cells and MS/MS-based identification. (A) Identification of deglycopeptides from both SKOV-3 and OVCAR-3 cells. (B) Identification of N-deglycopeptides by the GenoGlyco and MS/ MS-based methods. 5614

dx.doi.org/10.1021/pr400575f | J. Proteome Res. 2013, 12, 5609−5615

Journal of Proteome Research

■

Article

(15) Perkins, D. N.; Pappin, D. J. C.; Creasy, D. M.; Cottrell, J. S. Probability-based protein identification by searching sequence databases using mass spectrometry data. Electrophoresis 1999, 20, 3551− 3567. (16) Smith, R. D.; Anderson, G. A.; Lipton, M. S.; Pasa-Tolic, L.; Shen, Y.; Conrads, T. P.; Veenstra, T. D.; Udseth, H. R. An accurate mass tag strategy for quantitative and high-throughput proteome measurements. Proteomics 2002, 2, 513−523. (17) Zimmer, J. S. D.; Monroe, M. E.; Qian, W.-J.; Smith, R. D. Advances in proteomics data analysis and display using an accurate mass and time tag approach. Mass Spectrom. Rev. 2006, 25, 450−482. (18) Zhang, H.; Li, X.-j.; Martin, D. B.; Aebersold, R. Identification and quantification of N-linked glycoproteins using hydrazide chemistry, stable isotope labeling and mass spectrometry. Nat. Biotechnol. 2003, 21, 660−666. (19) Kaji, H.; Saito, H.; Yamauchi, Y.; Shinkawa, T.; Taoka, M.; Hirabayashi, J.; Kasai, K.-i.; Takahashi, N.; Isobe, T. Lectin affinity capture, isotope-coded tagging and mass spectrometry to identify Nlinked glycoproteins. Nat. Biotechnol. 2003, 21, 667−672. (20) Bause, E. Structural requirements of N-glycosylation of proteins. Studies with proline peptides as conformational probes. Biochem. J. 1983, 209, 331−336. (21) Zhang, H.; Loriaux, P.; Eng, J.; Campbell, D.; Keller, A.; Moss, P.; Bonneau, R.; Zhang, N.; Zhou, Y.; Wollscheid, B.; Cooke, K.; Yi, E. C.; Lee, H.; Peskind, E. R.; Zhang, J.; D Smith, R.; Aebersold, R. UniPep - a database for human N-linked glycosites: a resource for biomarker discovery. Genome Biol. 2006, 7, R73. (22) Ong, S.-E.; Mittler, G.; Mann, M. Identifying and quantifying in vivo methylation sites by heavy methyl SILAC. Nat. Meth. 2004, 1, 119−126. (23) Geiger, T.; Cox, J.; Ostasiewicz, P.; Wisniewski, J. R.; Mann, M. Super-SILAC mix for quantitative proteomics of human tumor tissue. Nat. Meth. 2010, 7, 383−385. (24) Monetti, M.; Nagaraj, N.; Sharma, K.; Mann, M. Large-scale phosphosite quantification in tissues by a spike-in SILAC method. Nat. Meth. 2011, 8, 655−658. (25) Boersema, P. J.; Geiger, T.; Wiśniewski, J. R.; Mann, M. Quantification of the N-glycosylated secretome by super-SILAC during breast cancer progression and in human blood samples. Mol. Cell. Proteomics 2013, 12, 158−171. (26) Tian, Y.; Zhou, Y.; Elliott, S.; Aebersold, R.; Zhang, H. Solidphase extraction of N-linked glycopeptides. Nat. Protoc. 2007, 2, 334− 339. (27) Zeng, Y.; Ramya, T. N. C.; Dirksen, A.; Dawson, P. E.; Paulson, J. C. High-efficiency labeling of sialylated glycoproteins on living cells. Nat. Meth. 2009, 6, 207−209. (28) Shankavaram, U. T.; Reinhold, W. C.; Nishizuka, S.; Major, S.; Morita, D.; Chary, K. K.; Reimers, M. A.; Scherf, U.; Kahn, A.; Dolginow, D.; Cossman, J.; Kaldjian, E. P.; Scudiero, D. A.; Petricoin, E.; Liotta, L.; Lee, J. K.; Weinstein, J. N. Transcript and protein expression profiles of the NCI-60 cancer cell panel: an integromic microarray study. Mol. Cancer Ther. 2007, 6, 820−832. (29) Kersey, P. J.; Duarte, J.; Williams, A.; Karavidopoulou, Y.; Birney, E.; Apweiler, R. The International Protein Index: An integrated database for proteomics experiments. Proteomics 2004, 4, 1985−1988. (30) Yamasaki, R. B.; Osuga, D. T.; Feeney, R. E. Periodate oxidation of methionine in proteins. Anal. Biochem. 1982, 126, 183−189. (31) Pruitt, K. D.; Tatusova, T.; Maglott, D. R. NCBI reference sequences (RefSeq): a curated non-redundant sequence database of genomes, transcripts and proteins. Nucleic Acids Res. 2007, 35, D61− D65.

ASSOCIATED CONTENT

S Supporting Information *

This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Phone: (410) 502-8149. Fax: 443-287-6388. E-mail: [email protected]. Author Contributions #

S.S. and B.Z. contributed equally to this work.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work is partially supported by National Institutes of Health, National Cancer Institute, Clinical Proteomic Tumor Analysis Consortium (U24CA160036), The Early Detection Research Network (EDRN, U01CA152813), R01CA112314, and Chia Family Foundation.

■

REFERENCES

(1) Rudd, P. M.; Elliott, T.; Cresswell, P.; Wilson, I. A.; Dwek, R. A. Glycosylation and the immune system. Science 2001, 291, 2370−2376. (2) Lis, H.; Sharon, N. Protein glycosylation. Eur. J. Biochem. 1993, 218, 1−27. (3) Tian, Y.; Zhang, H. Glycoproteomics and clinical applications. Proteom. Clin. Appl. 2010, 4, 124−132. (4) Haltiwanger, R. S.; Lowe, J. B. Role of glycosylation in develooment. Annu. Rev. Biochem. 2004, 73, 491−537. (5) Sun, S.; Wang, Q.; Zhao, F.; Chen, W.; Li, Z. Glycosylation site alteration in the evolution of influenza A (H1N1) viruses. PLoS One 2011, 6, No. e22844. (6) Ohtsubo, K.; Marth, J. D. Glycosylation in cellular mechanisms of health and disease. Cell 2006, 126, 855−867. (7) Durand, G.; Seta, N. Protein glycosylation and diseases: Blood and urinary oligosaccharides as markers for diagnosis and therapeutic monitoring. Clin. Chem. 2000, 46, 795−805. (8) Aebersold, R.; Mann, M. Mass spectrometry-based proteomics. Nature 2003, 422, 198−207. (9) Link, A. J.; Eng, J.; Schieltz, D. M.; Carmack, E.; Mize, G. J.; Morris, D. R.; Garvik, B. M.; Yates, J. R. Direct analysis of protein complexes using mass spectrometry. Nat. Biotechnol. 1999, 17, 676− 682. (10) Gygi, S. P.; Rist, B.; Gerber, S. A.; Turecek, F.; Gelb, M. H.; Aebersold, R. Quantitative analysis of complex protein mixtures using isotope-coded affinity tags. Nat. Biotechnol. 1999, 17, 994−999. (11) Ross, P. L.; Huang, Y. N.; Marchese, J. N.; Williamson, B.; Parker, K.; Hattan, S.; Khainovski, N.; Pillai, S.; Dey, S.; Daniels, S.; Purkayastha, S.; Juhasz, P.; Martin, S.; Bartlet-Jones, M.; He, F.; Jacobson, A.; Pappin, D. J. Multiplexed protein quantitation in Saccharomyces cerevisiae using amine-reactive isobaric tagging reagents. Mol. Cell. Proteomics 2004, 3, 1154−1169. (12) Ong, S.-E.; Blagoev, B.; Kratchmarova, I.; Kristensen, D. B.; Steen, H.; Pandey, A.; Mann, M. Stable isotope labeling by amino acids in cell culture, SILAC, as a simple and accurate approach to expression proteomics. Mol. Cell. Proteomics 2002, 1, 376−386. (13) Eng, J.; McCormack, A.; Yates, J. An approach to correlate tandem mass spectral data of peptides with amino acid sequences in a protein database. J. Am. Soc. Mass Spectrom. 1994, 5, 976−989. (14) Li, X.-j.; Zhang, H.; Ranish, J. A.; Aebersold, R. Automated statistical analysis of protein abundance ratios from data generated by stable-isotope dilution and tandem mass spectrometry. Anal. Chem. 2003, 75, 6648−6657. 5615

dx.doi.org/10.1021/pr400575f | J. Proteome Res. 2013, 12, 5609−5615